近年來，人們大量在網路上發表自己的想法及言論，因考慮到這些大量網路言論的價值，部份學者開始對這些資料做情感分析，若是能將情感分析和政治領域做結合，就能在短時間內得到民眾對候選人的支持率或是施政的滿意度，節省傳統問卷的時間及人力。然而，因情感分析資料來源來自於網路，搜集到的詞語勢必充斥著大量的網路用語，這些網路用語也含有強弱不等的情感在內，若將這些非正規情感詞排除在研究範圍之外，將會損失大量有價值的資料，因此，建立一個網路用語情感辭典，對以網路言論做情感分析有重大的作用。本研究蒐集2016總統大選前一個月的Facebook文章，分析各候選人的支持率，結果顯示在Facebook上以情感分析挖掘網路社群媒體民眾意見較趨向特定族群，和實際支持率有落差，。網路用語情感辭典目前因網路文章不嚴謹、結構不完整問題受到了很大的限制，但在一些限制條件下，正確率有微幅提升。而辭典的建制雖然還不是很完整，但也為將來分析網路文章時所需的領域專用情感辭典，做為一個開端。;In recent years, people express their opinions on the Internet frequently. Those internet comments is very valuable, then some researchers analysis those data by sentiment analysis. If sentiment analysis can be applied to politics, the approval rating of candidates will be gotten more efficient. However, the new words form Internet were not supported by existing sentiment dictionary. In this study, the practicability of applying sentiment analysis to politics illustrated is verified by the case of Taiwan 2016 presidential election on Facebook. In addition, a Internet slang words sentiment dictionary has been created for improving the accuracy of sentiment analysis. At last, there′s a gap from originally expected. The experiment result is tendency to a specific group when using the opinions on Facebook. The effectiveness of the new dictionary is limited because the grammar of the web article is not rigorous. But the accuracy is improved under special condition.
